Individuals with ASD display a unique pattern of abilities and disabilities in learning and memory. They exhibit intact lower-level learning of item, facts, details, and routines. However, they manifest impairments in higher- level learning involving abstraction, problem solving, using information that must be organized with self- generated conceptual schemas, and generalizing learning. This uneven and situation-focused learning profile has a profound impact on the academic, social, and adaptive functioning of those with ASD. Furthermore, while consistent with other neurocognitive theories of ASD, a learning-based approach offers clearer links to animal and computational model systems, thus enhancing its potential to illuminate pathophysiological mechanisms. In this application, I attempt to clarify the neural mechanisms underlying higher-level learning deficits, and help identify their precise effects on the day-to-day functioning of adolescents with ASD. Experiments will include behavioral and fMRI studies in a well-characterized group of adolescents aged 12-17 years, 11 months with ASD (n=40), and age, IQ, and gender-matched control participants with typical development (TYP; n=40).The first Aim, investigates the neural circuitry underlying higher-level learning deficits in adolescents with ASD using fMRI and a transitive inference paradigm which includes testing on lower-level trained and higher-level abstracted relationships between stimuli. This Aim employs a rapid event- related fMRI study which is interpreted in the context of our recent behavioral study of transitive inference learning in young adults, and two prominent mechanistic learning models. We predict that individuals with ASD will demonstrate reduced frontal functioning and functional connectivity during higher-level inferences. In the second Aim, we investigate how behavioral and neural indices of lower and higher-level learning relate to academic performance, social problem solving, and restricted and repetitive behaviors using regression analyses to examine how reading, writing, math, and social problem solving abilities, and restricted interests and repetitive behaviors, are predicted by neurocogntive measures of abstract reasoning and generalization, and neural indices of transitive inference. This work proposes a new mechanistic model of ASD; attempts to link neurocognition to real life problems; focuses on the understudied adolescent population during a potentially effective window for intervention; advances the search for psychopharmacological, psychosocial, and neural retraining remediation strategies for these individuals who are extremely able, but experience higher-level learning problems that limit their human potential; uses Bayesian state-space quantitative methods to closely examine the emergence of learning; and establishes a conceptual basis for future investigation of generalization of learning in mice through the development of parallel tasks in mouse models of ASD that can be used as preclinical assays for dopaminergic and other pharmacological treatments for cognitive deficits. PUBLIC HEALTH RELEVANCE: This study of adolescents with ASD and typical development uses fMRI and neurocognitive measures to examine the impact of higher level learning deficits and their effect on academic performance, social problem solving, and restricted interests and repetitive behaviors.